Pulse of the city:
Visualizing Urban Dynamics of Special Events
Andrea Vaccari
Mauro Martino
Francisca Rojas
Carlo Ratti
Massachusetts Institute of Massachusetts Institute of Massachusetts Institute of Massachusetts Institute of
Technology
Technology
Technology
Technology
77 Massachusetts Avenue 77 Massachusetts Avenue 77 Massachusetts Avenue 77 Massachusetts Avenue
Cambridge, MA 02139 USA Cambridge, MA 02139 USA Cambridge, MA 02139 USA Cambridge, MA 02139 USA
++ 1-617-324-5560
++ 1-617-324-5560
++ 1-617-324-5560
++ 1-617-324-5560
avaccari@mit.edu
mmartino@mit.edu
ABSTRACT
How does a city perform during a special event? How can we
inform local authorities and urban dwellers of public happenings?
Until today it has been difficult to monitor urban dynamics in real
time. Traditional methods, like head counts, surveys, and aerial
inspection, are costly and limited. In this paper we present our
twofold approach to the processing and visualization of real time
information gathered from the telecommunication infrastructures
that percolate modern cities.
fmr@mit.edu
ratti@mit.edu
and Behavioral Sciences]: Sociology.
General Terms
Design, Human Factors.
Keywords
Urban Dynamics, Urban Informatics, Urban Interaction Design,
Visualization for the Masses.
This paper presents visualizations for a popular audience that
examine the social psychogeography of the cityscape. Our work is
a step towards understanding the functions of the city through
digitally-enabled urban functions such as the pervasive
telecommunications network. Based on an analysis of data from a
telecom network, we illustrate how special events influence the
normal rhythms of the city, and how crowds move and respond to
large-scale public events.
This paper also discusses the potential of realizing an interface
aimed at informing urban dwellers in real time of the dynamics
that occur in the very place they find themselves. We contend that
this type of public intervention could enable local authorities,
service providers, businesses, and citizens themselves to function
and behave more efficiently, making the places they inhabit more
livable and sustainable.
This paper presents the two approaches, compares them to similar
recent projects, and discusses the design choices that drove our
implementation and the insights that we gained by their
implementation.
Categories and Subject Descriptors
H.5.1 [Multimedia Information Systems]: Animations; [J.4 Social
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies are
not made or distributed for profit or commercial advantage and that
copies bear this notice and the full citation on the first page. To copy
otherwise, or republish, to post on servers or to redistribute to lists,
requires prior specific permission and/or a fee.
GraphiCon'2010, September 20–24, 2010, St.Petersburg, Russia.
Copyright 2010 ACM 1-58113-000-0/00/0010…$10.00.
Fig. 1. Obama | One People [15] visualizes cell phone call
activity during the Inauguration of President Barack Obama
as a way to understand the urban dynamics of special events.
1. INTRODUCTION
Over the past decade there has been an explosion in the
deployment of pervasive systems like cell phone networks and
user-generated content aggregators on the Internet that produce
massive amounts of data as a by-product of their interaction with
users. This data is related to the actions of people and thereby to
the overall dynamics of cities, how they function and evolve over
time. Electronic logs of cell phone calls, subway rides, GPSenabled buses, and geotagged photographs are all digital
footprints [6] that today allow researchers to better understand
how people flow through urban space, and could ultimately help
those who manage and live in urban areas to configure more
liveable, sustainable, and efficient cities [2].
At the same time, these pervasive systems present the possibility
of extracting and inserting real-time information about social
dynamics into the built environment. The possibility of
establishing this dynamic feedback loop of information about how
a city functions has the potential to influence many aspects of
urban management by assisting local authorities, service
providers, enterprises, and even citizens themselves to make more
accurate decisions and thereby create a more efficient urban
environment. In order for this to happen, we contend that it is
necessary to first educate the public in understanding how
individual choices build up to form emerging urban processes that
affect the city and its inhabitants as a whole.
Obama | One People [15] employs digital footprints from a major
telecom operator’s cell phone network to examine how people
occupy urban space during a major special event. The project
consists of two visualizations, The City and The World, which are
aimed at a popular audience and represent the event from both a
local and an international perspective. We chose President Barack
Obama’s inauguration day in Washington D.C. as a case to
develop the special event visualizations and we partnered with a
major telecom operator to use cell phone data as proxy of social
activity. We designed and implemented the Obama | One People
visualizations to provide aesthetically compelling answers to
questions like: How does a city perform during a special event or,
analogously, a sudden emergency? Where do people congregate
and how does this vary over time?
It should be noted that there are many projects that focus on
specific aspects of urban pervasive systems and the study of
digital traces like cell phone calls, text messages and geotagged
photographs [6, 13, 7] and the implementation of new mobile
services [12, 1, 9] and interfaces [8, 14, 5]. Our approach in
Obama | One People is slightly different. The goal of our
visualizations is to translate the invisible digital traces of cell
phones and relate these to urban space so that people can grasp the
large-scale dynamics that occur within a cityscape. The ultimate
intent of this work is to envision a future where urban actors can
process existing information in real time directly from the city’s
digital infrastructure and make more informed decisions about
their social behavior within the city.
The data presented in Obama | One People consists of hourly
counts of cell phone calls per each base transceiver station
serviced by a major telecom operator in the Washington, D.C.
metro area along with the location where those phones are
registered, by state for American phones and by country for
international phones. The data set provided covered the period
from January 10, 2009 to January 28, 2009, out of which we
chose to represent the three days before and after the event, from
the 17 to the 23. To ensure the complete privacy of the telecom
provider’s customers, the data is in accordance with the provider’s
privacy policies and we performed our analyses in compliance
with the 2002 Directive of the European Parliament and Council
on Privacy. Our use of aggregate data implies that we cannot
identify individual callers, and we hope that our project might
stimulate a dialogue on the responsible access to digitallygenerated urban data and on how it can provide value-added
services to local and regional communities.
This paper first reviews the relevant design works in the field of
urban informatics, then discusses the design requirements we
sought to achieve in Obama | One People and describes the
visualizations themselves and their results, reviews the
architecture designed to implement the project, and finally
considers the limitations and practical applications of this type of
work. All in all, we believe Obama | One People offers the
following contributions to the field:
- It presents two visualizations that use existing digital
footprints to create an aesthetically compelling representation
of the temporal and spatial aspects of a public event in a major
urban area.
- It proposes how such visualizations can be accessed through
digital interfaces to support real-time decision-making in the
urban environment.
2. RELATED WORKS
Urban space and cartographic space are inseparable: from the grid
to the aerial perspective to contemporary digital mapping
technologies, the experience of urban space is always an
important component in people’s mental map of the city. The map
has become a popular interface in illustrating data sets drawn from
our increasingly digitally enabled urban infrastructures: from GPS
[18, 17, 4] to cell phone networks [13, 11] and other objects
equipped with radio frequency ID tags [4, 3]. The purposes behind
these types of visualizations vary: from purely artistic endeavors
to traffic monitoring to the uncovering of emerging patterns of
urban activity.
Fig. 2. Examples of visualizations of urban dynamics.
There seem to be four distinctive elements in the
visualization of urban digital footprints:
Infrastructure and scalability. The type of data being used
in the visualizations are generated by digitally-operated urban
infrastructures in the course of the everyday functions of the city
(RFID, WiFi hotspots, GPS, cell phones).
Flows and contexts. The data are displayed dynamically
over time and in geographic space in order to represent flows of
activity (i.e. pedestrians and cars). Some visualizations attempt to
represent activity in almost real-time, often with a slight time lag
for processing, such as a 15-minute delay.
2D and 3D perspective views. Depending on the objective
of the visualization, some perspectives are in 2D, such as in
representing traffic flows, and others are in 3D to present concepts
like the volume of activity in urban space.
Levels of data aggregation. The scale at which the data is
presented varies in detail and aggregation, such as in the case of
sensitive information about individuals.
The following is a discussion of precedent projects that exemplify
the most notable approaches to mapping urban dynamics and that
offer starting points for illustrating the city-wide special event for
Obama | One People.
Amsterdam Real Time (2003) by the Waag Society and Esther
Polak [18] (Figure 1.a) focused on visualizing the connection
between the structure of the city and the movements of
Amsterdam’s inhabitants.
Fig. 3. Distinctive elements in the visualization of urban digital
footprints.
By examining people’s mobility through GPS, the project
uncovered a map of the city defined solely by people’s activity in
space. Only GPS trajectories were recorded in the visualization
and this urban flux revealed Amsterdam’s urban form only by
capturing the digital traces of people’s movements (Figure 2.a).
Similar to Cabspotting in San Francisco [17] (Figure 1.b) this
visualization employs the flow information to structure the
geographic space of the city, highlighting thoroughfares of
movement but leaving out buildings and public spaces. Because of
a lack of geographic references, it is difficult for those not familiar
with the map of Amsterdam to understand the connection between
the form of the built environment and the flows of people.
Nevertheless (Figure 2.b), using GPS systems to visualize urban
dynamics is a powerful way to detail (Figure 2.d) the movements
of large numbers of people. Though as in all projects that use
GPS, extending this kind of mapping to include larger number of
individuals involves a high cost and a major organizational effort.
As mentioned previously, Cabspotting is similar to Amsterdam
Real Time [18] in visualizing movement within the city and in its
design approach. Directed by Scott Snibbe [17], Cabspotting uses
GPS technology Figure 2.a in San Francisco’s taxicabs to trace
their movement in real time through the city. The patterns traced
by each cab create a dynamic sense of traffic flows through the
city’s arteries, revealing where traffic is moving quickly and
where it is congested.
Real Time Rome [13] (Figure 1.c) and UrbanMobs [11] (Figure
1.d) introduce a 3D perspective view (Figure 2.c) and provide a
sense of the collective emotions of a city. As the MIT SENSEable
City Lab’s contribution to the 2006 Venice Biennale, Real Time
Rome took aggregated data from cell phones and mapped these
calls onto the geography of the city during two special events over
the summer of 2006: the World Cup finals match between Italy
and France, and a Madonna concert. The visualizations show
peaks in the volume of calls during stirring moments (such as
Italy scoring a goal during the World Cup match) in a sense
revealing the emotional signature of the city as well as where
people are congregating. Similarly, Orange Labs and Faber
Novel developed UrbanMobs as a tool to showcase popular
emotion cartography through the analysis and visualization of
citywide cellular network traffic activity. Both examples signal a
shift in both the aesthetic qualities of visualizing dynamic urban
data and in the methodology of positioning cell phones within
urban space according to the location of cell phone towers that
service those calls. This methodology may lose the detail of GPS
data but in tapping into the network infrastructure of cell phones,
it can harness vast amounts of data representing large swaths of
the city, thus increasing the scale of representation.
Measuring and mapping people’s emotional responses to the built
environment itself is the goal of BioMapping by Christian Nold
[4] (Figure 1.e). This is a community-mapping project that utilizes
a GSR (Galvanic Skin Response) device to measure how people
react psychologically to different areas of the city. Over 1,500
people have contributed to this exercise in various cities to
produce these communal emotion maps. Presented in a 3D view,
the visualizations use Google Earth to relate people’s emotional
arousal as measured by the GSR device to the geographic
locations they traverse. BioMapping clearly presents the urban
context of people’s reactions to the built environment, and the
GSR sensors and GPS scale the visualizations down to show
individuals’ responses; however the number of people involved
limits the extent to which these maps may present a collective
psychogeogaphy of the city.
Another project that visualizes data gathered directly from
consenting individuals is CamMobSense in Cambridge, UK [3]
(Figure 1.f). Using sensors mounted on pedestrians and cyclists,
CamMobSense monitors pollution in the city and relays the
collected data to a website in real time. It presents a low-cost,
distributed sensor model where people themselves collect data
that can be then visualized to better understand the environmental
impacts of urban functions such as transportation, and in turn this
can help those who manage urban regions make more informed
decisions.
Building upon these precedents, Obama | One People utilizes
aggregated data from cell phones and overlays the call dynamics
onto an aerial photograph of Washington D.C., which highlights
the event spaces of the Presidential Inauguration such as the Mall
and Pennsylvania Avenue. We present the visualization in 3D to
highlight the peaks and troughs in cell phone activity before,
during and after the special event itself, totalling a week’s time of
data. While the aggregated data does not reveal details of
individual movements, it does reveal collective dynamics over the
course of a week in large swaths of the Nation’s Capital. We have
thus created a scalable system that is able to capture the call
activity of large numbers of cell phones resulting in a
visualization that highlights the communications flows of a crowd
and how these relate to geographic locations within the city.
3. DESIGN REQUIREMENTS
In light of the above discussion, we identified the following
specific design requirements to guide the design of Obama | One
People. The visualizations should create a ”digital skin” of urban
spaces, with a simple and clear visual language system. They
should convey the social dynamics of a crowd, which are real
phenomena, using informational data. The user should perceive
the former, not the latter. They should establish a strong
relationship between the crowd and the urban landscape. For this
reason, we considered the spatial representation of the information
as a given, reflecting the geographic nature of the data.
The visualizations should be interactive but simple, allowing the
user to explore the evolution of the event in space and time.
However, they are not meant to be visual analytics tools, so they
should not provide quantitative details on the data. To achieve
this, there should be a good balance between aesthetics and
functionality. Inspired by Norman’s work [10], we should use
color, saturation, and luminosity to improve the aesthetic and
emotional impact of the project.
Finally, from a technical perspective, Obama | One People should
ideally run on a real-time stream of data. However, due
restrictions on the time required to collect the data from the
telecom provider’s network it was not possible for represent the
event as it occurred.
4. OBAMA | ONE PEOPLE
On January 20, 2009, Washington D.C. hosted the largest crowd
in its history for a very special event: the Inauguration of
President Barack Obama. This crowd presented great challenges
to both federal and municipal authorities in terms of management,
transportation, security, and emergency response. Where would
people congregate? How would they get there? When would they
arrive and depart? From where were they coming? The authorities
even had difficulty in estimating how many people would attend
the event. Early predictions called for over 3 million people, and it
was only after the week of the Inauguration that the City was able
to release an official estimate of 1.7 million people. The 2009
Presidential Inauguration offers perhaps one of the most extreme
cases of the value of gathering data about large crowds in real
time so that decision makers can understand and manage special
events as they occur.
In considering the volume of mobile phone calls at any given
momentas indicators of a crowd’s geographic distribution over
time, our access to aggregated cell phone data from a large
telecom operator presented us with the opportunity to explore the
possibilities of using such data to understand the dynamics of
large-scale special events such as the Presidential Inauguration.
The Obama | One People visualizations sought to explore such
questions as: Who was in Washington, D.C. for President
Obama’s Inauguration Day? When did they arrive, where did they
go, and how long did they stay? The nature of interpersonal
communications also allows us to interpret the resulting
visualizations of call data as revealing the psychogeography of the
city during a special event. Guy Debord, a French theorist,
defined psychogeography in 1955 as the ”the study of the precise
laws and specific effects of the geographical environment,
consciously organized or not, on the emotions and behavior of
individuals.” [TODO Guy Debord. 1955. Introduction to a
Critique of Urban Geography (essay)]
Thus, another question Obama | One People explores is: when
and where did the crowd in Washington D.C. sense the need to
share thoughts, information, and feeling with others who were not
co-present? Not only does Obama | One People map the invisible
flows of communications over the geography of the city, it also
begins to gauge the collective emotional pulse of the city at a
large scale and over time. Below we discuss the two visualizations
produced to present our analyses of aggregated phone calls in
Washington D.C. during the week of President Barack Obama’s
Inauguration: The City and The World.
4.1 The City
The City illustrates two pieces of information about the
Presidential Inauguration: 1) where the crowd came from and 2)
the emotional flow of the massive event in Washington, D.C.
The City summarizes this information by relating call activity to
the geography of Washington D.C. We overlay the map of the
Nation’s Capital with a 3D color-coded animated surface of
square tiles, with one tile representing a geographic area of 150 x
150 meters. Each tile rises and turns red as call activity increases
and likewise drops and turns yellow as call activity decreases. On
the left side of the screen, a dynamic bar chart breaks down the
call activity by showing the normalized contributions of calls
from the U.S.’s fifty states and 138 foreign countries grouped by
continent. The timeline at the bottom illustrates the weeklong
trend of call activity in Washington, D.C., which follows the 3D
square tiles that rise and fall on top of the map of the city.
In visualizing the volume of calls served by a large telecom
network in Washington DC, it is possible to see peaks of call
activity as the crowd anticipates Obama’s oath, a drop in call
activity as the crowd listens to Obama’s address, and peaks again
as the crowd celebrates the inauguration of the new President.
Through their cell phones, those present at the historic event share
their impressions with friends and family in vast numbers: on the
morning of January 20th, call activity is two to three times
stronger than usual, and it rises to five times the normal levels
after 2 pm as President Obama takes his oath and people begin to
celebrate.
Fig. 4 The City visualizes call activity in the week of the
inauguration. Superimposed on the map of Washington, D.C.
is a 3-D color-coded animated surface of square tiles (1 tile
represents an area of 150 x 150 meters). Each tile raises and
turns red as call activity increases and likewise drops and
turns yellow as activity decreases. On the left, a bar chart
breaks down the call activity by showing the normalized
contributions of calls from the 50 states and 138 foreign
countries grouped by continent. The timeline at the bottom
illustrates the overall trends of call activity in the federal areas
of Washington, D.C., which are represented with 3-D yellow
models on the map at the center of the screen.
4.2 The World
The World reveals the international nature of Inauguration Day.
By using the area codes and country codes of the cell phones
present in Washington D.C. during the week of the Inauguration,
this visualization traces the trajectories of people travelling from
all over the U.S. and the world. We interpret the variations in call
activity as flows of people arriving in Washington, D.C. and then
departing the capital to go back home. The visualization employs
a world map to link the city of Washington to US states and
countries abroad. Packets of information representing 100 calls
for US states and 10 calls for foreign countries move to and from
Washington depending on whether call activity increased or
decreased in relation to the previous hour. The timeline on the
bottom of the screen shows the overall trend of call activity in
Washington, D.C., relating the flows of people with events in the
city.
This visualization shows that people from almost every corner of
the world and almost all fifty states attended the Inauguration of
President Barack Obama. The aggregated call data indicates that,
at the least, there were people present from 138 countries, totaling
over half of all the countries in the world. This is likely a low
estimate since the data represents only one domestic telecom
carrier. Among the foreign countries represented, the main
international callers are from Canada, Great Britain, France, and
Puerto Rico, which register a five-fold increase in call activity. In
the U.S., the top calling states are also the country’s most
populous: California, Florida, New York, and Texas. Notably,
Georgia also figures in the list of top five callers on Inauguration
Day, even though it ranks ninth in US population.
4.3 Implementation Choices
In the previous section we described the design requirements of
the two visualizations. Here we explain the thinking that
motivated our implementation choices with respect to the those
requirements.
First, we chose to represent the data from two perspectives. The
City shows the evolution of the event within the context of
Washington D.C. and allows users to see the build up in
anticipation of the Inauguration, and how the crowd faded out
after its conclusion. The World represents the international nature
of the event and aims to convey the idea of a global pilgrimage to
the American capital by displaying a map of the world and
showing how people from near and far gathered in one city for
one reason. In both visualizations, one aim is to represent the
social dynamics that enveloped the city before, during, and after
the President’s Inauguration. Another aim is to help users
understand how participation in a public event, like other
individual choices in the context of the urban environment, builds
up to create a collective social dynamic that affects the city as a
whole. For this second reason, we purposefully deemphasized the
quantitative nature of the data by using the visual metaphors of
waves and packets.
In formulating how the data representation would occur, we
considered more abstract visualizations of the flow of phone calls
but instead decided to use maps given the inherently geographic
nature of the data and the message we aim to convey about the
influence of special events on urban space. Moreover, we employ
map details to assist users in orienting themselves within the
context of Washington D.C. without overloading the visualization
with unnecessary details. For this reason, The City visualization
highlights only the major roads of Washington D.C. and the
Potomac River, while The World simplifies the representation of
flows by showing only connections between Washington D.C.,
US state capitals and the countries around the world.
While we simplified the representation Washington’s geography,
we decided to represent call activity in The City with some
redundancy along different features of the visualization: we used
color, opacity, and position of the tiles to represent the intensity of
activity within each cell. The combination of these three features
makes for a strong representation. Color varies from unsaturated
yellow to saturated red to represent the volume of activity (a
feature sensitive to color blind users). The use of opacity hides
those tiles where activity is low and highlights those where it is
strong, thus guiding the user through the evolution of the event.
Finally, the third dimension recreates the wave metaphor
mentioned above and further helps to convey the idea of the city
as a living system.
In The World visualization we represented the origin of the
callers’ phones only using packets of information. In this case,
employing multiple features as in The City would have focused
attention on those connections with the strongest activity, whereas
in this case we sought to highlight the overall exchange of inflows
and outflows happening in Washington D.C. For this, identical
packets are a strong solution as their movements alone create
denser or thinner flows emerging from Washington D.C.
The timelines placed at the bottom of the screen unify the two
visualizations and highlight the weeklong trend of activity in the
areas of the Washington D.C. where the Inaugural events took
place. This common element is intended to help the user associate
the information represented in one visualization with that being
represented in the other. The two visualizations also share a
minimalist interaction system that allow the user to navigate the
3D space with the mouse, using left and right clicks to rotate and
span, and the mouse wheel to zoom in and out. Keyboard arrows
can be used to pause and play the animation, and to move
backward and forward in time.
4.4 Results
Fig. 5. The World visualizes call traffic to and from the
capital. Variations in call activity are represented here as
flows of people coming to Washington, D.C. and then leaving
the capital to go back to their home states and countries. The
world map links Washington, D.C. to capitals abroad. Packets
move to and from Washington depending on whether call
activity increased or decreased in relation to the previous
hour. The timeline on the bottom of the screen shows the
overall trends of call activity in the federal areas of
Washington, D.C., allowing to associate the flows of people
with the events highlighted in The City.
The visualizations of aggregated call activity during the
Presidential Inauguration we created for Obama | One People
have provided insights into the dynamics of a remarkable special
event in Washington D.C. in terms of the composition of the
crowd and its emotional behavior in space and over time.
By examining the relative increase in call activity during the week
of the Inauguration as compared to a normal week of call activity
in the Nation’s Capital, it is possible to get a sense of who
travelled to Washington to attend this special event. The states
with the strongest increase in calls were the southern states of
Alabama, Georgia, Kentucky and Tennessee, with calls up to
twelve times the normal levels. These are states that played a
prominent role in the Civil Rights movement and notably are also
red states whose voting population went for the Republican
candidate, John McCain. Other states with a ten-fold increase in
call activity were Illinois, Barack Obama’s home state, and
Michigan, Ohio and Indiana, swing states which went blue, voting
for President Obama. These results illustrate that participation in
the event from each U.S. state did not depend on its political
leaning or geographical proximity to Washington, D.C.
As the visualizations show, call activity in the days before and
after January 20 reveal that the Inauguration was a multi-day
event as traffic increased markedly throughout the week. On
Sunday, January 18th call activity began to increase and, further
analysis reveals that on Monday it was already two to three times
what it would normally be on a Monday in January in Washington
D.C. On the day of the Inauguration, call activity reached
unprecedented levels throughout the day and the telecom
provider’s network served six times more SMS than normal
levels. In the morning, call activity was two to three times
stronger than normal levels as the crowd anticipated Barack
Obama’s oath, then dropped as the crowd listened to President
Obama’s inaugural address and peaked to five times the normal
levels after 2 pm as President Obama took the oath and people
celebrated.
Geographically, The City visualization reveal how the Inaugural
visitors spread beyond the confines of the Federal areas of
Washington, D.C. - such as the Mall and Pennsylvania Avenue into the neighborhoods of the city. The hotspots of activity were
clustered in the Northwest neighborhoods of the city, around
Downtown, Adams Morgan and U Street. The historic
inauguration of President Obama offered people from all over the
United States and the world the occasion for a truly urban
celebration. While these results are not surprising in themselves,
our analyses allow for the first time to quantify and compare the
presence of people in different areas of the city and to reveal their
dynamic movements and, to a certain extent, their emotional
reactions through time.
In post-project discussions with the telecom provider we learned
that while the origin of the crowd is not a key factor driving
capacity planning decisions of network operators, the remaining
questions are relevant. The visualization capabilities presented in
this paper made it possible to understand that the Inauguration
was a multi-day event that began two days before January 20th
and which did not end immediately, stretching visitors’ presence
in the Nation’s Capital through the middle of the week. The
telecom operator appreciated the visualizations’ simplicity and
their ability to summarize vast amounts of data, both temporal and
geographical. Obama | One People also convinced network
engineers of the benefits of collecting, analyzing and visualizing
digital footprints of telecommunications flows. This in turn led to
a new partnership with the radio access network engineers to get
access to more data and to get the support of their management for
further analysis.
Future user studies should include those who manage municipal
services during special events who often must make educated
guesses about crowd capacity in the city’s public spaces.
Moreover, it would be valuable to gauge how the information
transmitted through these visualizations could alter the behavioral
decisions of the participants of special events themselves. Would
they decide to arrive at the event earlier or later? Would they
choose to reorganize themselves in the public spaces of the as
they learn the real time dynamics of the crowd? Ultimately, what
these visualizations aim to do is to create a feedback loop that
provides real time information about urban dynamics back to
people so that they can adjust their decision-making as they go
about their daily activities.
Fig. 6. Trends of normalized call activitity before, during, and
after the event.
5. SYSTEM IMPLEMENTATION
We developed Obama | One People on top of an agile architecture
that supports the operations of data gathering, storage, and
analysis. These three operations exist in three separate layers and
we implemented the system in the Python programming language,
using the open-source PosgreSQL database system and the
PostGIS geospatial extension. Figure X details the workflow
between the layers, which we describe below.
5.1 Data Gathering
This layer provides standard tools for uploading urban data. It
consists of a secure FTP server and a server daemon. The server
allows the telecom operator to upload CSV files containing data
on the network activity in the last 60 minutes. The daemon detects
when a new file has been loaded, then parses it and loads the data
in the Data Storage layer.
We chose not to implement more advanced data streaming
technologies to simplify the interaction with the telecom operator.
This solution allowed us to obtain data in different formats and
resolutions and to rapidly restructure it in a form that is
compatible with the data storage layer. However, this solution did
not allow us to visualize network activity in real time.
Nevertheless, we were able to access the data and complete the
visualizations within a couple of weeks of the Inaugural event.
5.2 Data Storage
This layer stores the network activity data into a geospatialenabled database, which also stores information about the location
of the base transceiver stations and on the geographical structure
of the spatial grid used in The City visualization. Each data point
has a time reference and a geographic reference to the country or
state where the phones making calls are registered, which allows
inter-country normalization and continent-scale aggregation of the
data.
Since in this dataset the activity generated by U.S. phones is on
average two orders of magnitude higher than that generated by
foreign phones, we chose to normalize all time series in the range
[0; 1]. To do this, we divided the activity by the maximum
reached by each state or country within the time span of our data
set. In most cases, the maximum was right after President
Obama’s Inaugural speech (see Figure 6).
5.3 Data Analysis
This layer accesses data from the Data Storage layer and performs
geospatial operations that redistribute the punctual information on
network activity over the two-dimensional Washington, D.C. area.
The Data Analysis layer is also in charge of performing preprocessing operations on the data, like correcting missing values
and normalization. The geospatial operation adopted to
redistribute the network activity visualized in The City was based
on a discrete grid of 150x150 meter cells that cover the
Washington, D.C. area. The activity in each cell is estimated by
summing the contributions of all the base transceiver stations
weighted according to the distance between the cell and the
station. The weighting factor is defined by the exponential decay
function w=exp((dist(cell,bts)2)/0.2) which causes the weighting
factor to become insignificant at the reasonable distance of 1 km
from the station. The operation chosen to estimate the flows
visualized in The World was based on the gradient of the network
activity generated by each US state or country. Under the
assumption that the average number of calls per person does not
change radically over the course of the special event, we
estimated a larger presence of people from a given state or
country when the related call activity increased from one hour to
the next. To account for normal daily fluctuations, we subtracted
the average variations and represented only unusual fluctuations.
The main objective of our analyses was to summarize the data
collected by the cell phone network into two easily
understandable metaphors of social dynamics and urban
communications flows. We consciously avoided using more
invasive analyses, for example using radio wave propagation
models to redistribute the data, in favor of an approach that
focuses the viewers’ attention on the high-level understanding
about urban processes that this data can provide.
6. LIMITATIONS
Up to this point we have not able to validate the efficacy of the
visualizations as the project is still confidential and cannot be
shown outside our research group (we have specific permission
for this publication). In this section we describe the major insights
and limitations of our approach, which will be addressed in a
future user study. Thus far, we have been able to collect informal
feedback from the research scientists of the telecom operator,
which supports our design and implementation choices. Their
positive feedback confirms that the Obama | One People
visualizations offer a new perspective on the nature and dynamics
of a special event such as Inauguration Day. While the main
results are not surprising - it is to be expected that call activity
would increase markedly on the day of the event – they
nevertheless reinforce assumptions about how people may behave
during large-scale urban events. As service providers, the telecom
operator valued the insight that there was a slow build-up to the
actual day of the Inauguration and likewise a slow retreat of the
crowd from the city over the course of several days. This may
present a future scenario where the provider deploys extra telecom
capacity for a special event a few days before the event and leaves
it up and running for longer than expected.
Moreover, while most of the operator’s feedback praised the
simplicity of the interface and the visual system, some would have
liked to see more analytical tools included in the visualization so
that users could themselves examine the data. In the future,
therefore, we intend to create an interface that allows the experts
to study more in detail the data while maintaining a simple
interface for basic users. On the technical side, we limited our
visualizations to cell phone activity recognizing that there are
other forms of digital footprints. Works that used vehicle traffic
data, for example, have been discussed in the section on Related
Works. While it is not clear which datum or combination of data
is the best proxy for estimating social activity, our visualizations
could be immediately used with traffic data. Again, however, the
focus of our project is not on quantitative precision but rather on
the qualitative narrative.
Finally the architecture implemented is quite simple. While this
allows designers to easily adapt to changes in format and
resolution of the data, it does not allow the gathering, storage, and
analysis of data in real time. This is a major drawback that is due
to the slow process of data collection on the network side by the
telecom operator. In the future, however, as information will flow
more freely, a more advanced system architecture that fully
supports real time data streams will be necessary. In that scenario,
the simple geospatial procedures on the data that we perform will
not clog the system.
7. DISCUSSION
Obama | One People is part of a long term vision centered on the
idea that diffused social consciousness regarding urban dynamics
can improve the efficiency of cities by enabling local authorities,
service providers, businesses, and citizens to make better
informed decision in their urban landscape. This vision was
presented in WikiCity [16] (see Figure 7) that considered another
public event, the Notte Bianca in Rome, Italy, an all-night festival
that hosts artistic performances at various locations throughout the
city. In that occasion, we developed a public interface that
featured a satellite image of Rome with four types of information
overlayed on top of it: real-time cell phone activity, the real-time
location of public transport buses, starting and ongoing event tags
at their corresponding location, and live news feeds from
journalists at key locations. While the scope of this paper is not to
present WikiCity, we believe it is important to discuss its
underlying concept to contextualize Obama | One People.
Through WikiCity we envision the participatory provisioning of
urban information by local authorities, service providers,
businesses, and citizens, and the enabling of semantic processes
for the development of an evolving hierarchy of analysis tools and
information visualizations. These mechanisms, coupled with
ubiquitous interfaces, are expected to enact the active use of
digital technologies for real time location-based decision making
in the city. In a similar fashion, they are intended to support the
active collection of public feedback on the condition of the city
and its services. The mapping of urban information is not limited
to providing a real time representation of the city. Instead it
becomes an instrument for citizens and local authorities to base
their actions and decisions in a better informed manner. In this
way the real time information changes the city context as well as
is changed by that altered context accordingly. The potential in
this scenario is to address a wide variety of problems that interest
the multifaceted aspects of managing the city as a complex
system. The results of WikiCity were mixed. If on one side it
proved the technical feasibility of our vision, on the other it failed
in creating a user base and the visualization was not adopted for
subsequent major public events. While there are many possible
reasons behind this, we believe that one of the major issues was
that our interface tried to convey too much information to a public
that was almost completely unaware of the significance of that
information. We see Obama | One People as an intermediate step
in dramatically illustrating and popularizing the idea of the city as
a living system that responds collectively to the actions and
decisions of individuals.
8. CONCLUSION
This paper discussed a visualization project that considered how
crowds behave during large-scale public events in urban areas.
Obama | One People took as its case the Inauguration of President
Barack Obama and overlaid telecommunications data over the
geography of Washington D.C. The two resulting visualizations –
The City and The World – showed from a local and an
international perspective that the Inauguration was a multi-day
event and that it drew people from all over the US and the world.
The potential of visualizations such as Obama | One People are
great for those who manage and act within urban space. They are
a first step in revealing collective patterns of social dynamics, and
data such as telephone call activity can offer a insights into how
people inhabit the city in different times and locations, and how is
the urban environment can be stressed in the case of public events
or sudden emergencies.
Having the results of these analyses in real-time, urban planners
would be able to promptly detect and correct phenomena that
reduce the liveability and sustainability of the city: instead of
planning urban interventions and waiting months to evaluate their
impact, they could switch to a more reactive, real-time
management of the city. Moreover, these kinds of visualizations,
if accessed through an appropriate interface, can also be of value
to urban dwellers themselves. The ultimate goal of this work is to
build urban interfaces where people can deposit and extract
information on the functions of the city in real time and thus work
together in building more efficient, intelligent and sustainable
cities.
[2] F. Calabrese, K. Kloeckl, and C. Ratti. Handbook of
Research on Urban Informatics, chapter WikiCity: RealTime Location-Sensitive Tools for the City. IGI Global,
2008.
[3] Centre for Scientific Computing, University of Cambridge in
collaboration whit Imperial College London, Leeds
University, Newcastle University and Southampton
University. CamMobSens,
http://www.escience.cam.ac.uk/mobiledata, 20 February
2010.
[4] Christian Nold. Bio Mapping,
http://biomapping.net/index.htm, 15 February 2010.
[5] S. Eisenman, N. Lane, E. Miluzzo, R. Peterson, G. Ahn, and
A. Campbell. Metrosense project - people-centric sensing at
scalAe.CM Conference on Embedded Networked Sensor
System, 2s006.
[6] F. Girardin, F. Calabrese, F. Dal Fiore, C. Ratti, and J. Blat.
Digital footprinting: Uncovering tourists with user-generated
contenIEt.EE Pervasive Computing Magazine, 7(4):36–43,
2008.
[7] M. Gonzalez, C. Hidalgo, and A. Barabasi. Understanding
individual human mobility patterns.Nature Magazine, 2008.
[8] . Liu and F. Zhao. Towards semantic services for sensor-rich
information systems.2nd IEEE/CreateNet
InternationalWorkshop on Broadband Advanced Sensor
Networks , 2005.
[9] R. Murty, A. Gosain, M. Tierney, A. Brody, A. Fahad, J.
Bers, and M. Welsh. Citysense - an urban-scale wireless
networking testbed. IEEE Conference on Technologies for
Homeland Security, 2007.
[10] D. Norman. Emotional Design: Why we love (or hate)
everyday things. Basic Books, 2004.
[11] Orange Labs and Faber Nove. UrbanMobs,
http://www.fabernovel.com, Revised in 22 April 2010.
[12] A. Pawling, T. Schoenharl, P. Yan, and G. Madey. Wiper:
An emergency response system. 6th International Conference
on Information Systems for Crisis Response and
Management, 2008.
[13] J. Reades, F. Calabrese, A. Sevtsuk, and C. Ratti. Cellular
census: Explorations in urban data collection. IEEE
Pervasive Computing Magazine, 6(3):30–38, 2007.
[14] A. Santanche, S. Nath, J. Liu, B. Priyantha, and F. Zhao.
Senseweb - browsing the physical world in real time.
Microsoft Research, 2006.
[15] Senseable City Lab, Obama | One People,
http://senseable.mit.edu/obama, Revised in 6 May 2010.
Fig. 7. WikiCity [16] was publicly displayed during a major
event in Rome, Italy and presented real-time cell phone
activity, the real-time location of public transport buses,
starting and ongoing event tags at their corresponding
location, and live news feeds from journalists at key locations.
9. REFERENCES
[1] M. Arikawa, S. Konomi, and K. Ohnishi. NAVITIME:
Supporting pedestrian navigation in the real world.IEEE
Pervasive Computing Magazine, 2007.
[16] Senseable City Lab, WikiCity,
http://senseable.mit.edu/wikicity, Revised in 6 May 2010.
[17] Stamen Design, Produced by Peter Richards and Susan
Schwartzenberg, directed by Scott Snibbe. Cabspotting,
http://cabspotting.org, Revised in 20 May 2010.
[18] Waag Society together with Esther Polak. Amsterdam
RealTime, http://senseable.mit.edu/wikicity/, Revised in 15
April 2010.